discover-event-surfaces
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from codebase files and user-supplied descriptions to map product flows.
- Ingestion points: The skill reads a
change_briefYAML input and performs full file reads of components and their dependencies in the codebase (SKILL.md). - Boundary markers: The instructions lack specific delimiters or guardrails to prevent instructions embedded within the analyzed source code from being interpreted as agent commands.
- Capability inventory: The skill utilizes file system read access and platform tool invocations for fetching project context (
get_context) and existing events (get_events). - Sanitization: There is no evidence of sanitization or filtering of codebase content to strip potentially malicious instructions before processing.
Audit Metadata